OPERATIONAL PLANNING IN COMBINED HEAT AND POWER SYSTEMS

This dissertation presents methodologies for operational planning in Combined Heat and Power (CHP) systems. The subject of experimentation is the University of Massachusetts CHP system, which is a 22 MWe/640 MBh system for a district energy application. Systems like this have complex energy flow networks due to multiple interconnected thermodynamic components like gas and steam turbines, boilers and heat recovery steam generators and also interconnection with centralized electric grids. In district energy applications, heat and power requirements vary over 24 hour periods (planning horizon) due to changing weather conditions, time-of-day factors and consumer requirements. System thermal performance is highly dependent on ambient temperature and operating load, because component performances are nonlinear functions of these parameters. Electric grid charges are much higher for on-peak than off-peak periods, on-site fuel choices vary in prices and cheaper fuels are available only in limited quantities. In order to operate such systems in energy efficient, cost effective and least polluting ways, optimal scheduling strategies need to be developed. For such problems, Mixed-Integer Nonlinear Programming (MINLP) formulations are proposed. Three problem formulations are of interest; energy optimization, cost optimization and emission optimization. Energy optimization reduces system fuel input based on component nonlinear efficiency characteristics. Cost optimization addresses price fluctuations between grid on-peak and off-peak periods and differences in on-site fuel prices. Emission optimization considers CO2 emission levels caused by direct utilization of fossil fuels on-site and indirect utilization when importing electricity from the grid. Three solution techniques are employed; a deterministic algorithm, a stochastic search and a heuristic approach. The deterministic algorithm is the classical branch-and-bound method. Numerical experimentation shows that as planning horizon size increases linearly, computer processing time for branch-and-bound increases exponentially. Also in the problem formulation, fuel availability limitations lead to nonlinear constraints for which branch-and-bound in unable to find integer solutions. A genetic algorithm is proposed in which genetic search is applied only on integer variables and gradient search is applied on continuous variables. This hybrid genetic algorithm finds more optimal solutions than branch-and-bound within reasonable computer processing time. The heuristic approach fixes integer values over the planning horizon based on constraint satisfaction. It then uses gradient search to find optimum continuous variable values. The heuristic approach finds more optimal solutions than the proposed genetic algorithm and requires very little computer processing time. A numerical study using actual system operation data shows optimal scheduling can improve system efficiency by 6%, reduce cost by 11% and emission by 14%